UC Irvine

This dissertation contributes towards cooperative localization (a form of peer-assistive localization/navigation solution) for networked mobile agents (robots, humans, unmanned aerial vehicles (UAV), etc.) in Global Positioning System (GPS) denied environments. The objective is the design of infrastructure-free decentralized cooperative localization solutions that can work in complex environments. The main driving application of this work is cooperative localization solutions for firefighters and first-responders in extreme indoor environments where taking inter-agent measurements in line-of-sight (LoS) and maintaining network connectivity are challenging. This dissertation work is end-to-end, spanning algorithm design, theoretical modeling and analysis, testbed development, and experimental demonstrations/validation.

At the algorithmic level, the main challenge in the design of decentralized cooperative localization is how to handle inter-agent correlations properly under limited connectivity. We propose novel loosely coupled cooperative localization solutions that relax the connectivity requirement and reduce the communication cost by avoiding the exact track of correlations but accounting for them in an implicit manner. In our first solution, to guarantee filter consistency, we account for unknown inter-agent correlations due to the past relative measurement updates by using a well-established discorrelated upper bound on the joint covariance matrix of the agents. In the second method, we use an optimization framework to estimate unknown inter-agent cross-covariance matrices. Rigorous analysis establishes the formal performance guarantees of our algorithms. At the implementation level, we use ultra-wideband (UWB) technology for both inter-agent infrastructure-free communication and inter-agent ranging sensors. UWB technology has high accuracy, high data rate, ability to take range measurements in non-line-of-sight (NLoS), and interference immunity comparing to other technologies that make it an attractive choice for complex environments. However, there are challenges that should be addressed to adopt this technology for our purpose. For UWB ranging, the ranging accuracy degrades substantially due to positive bias in NLoS measurements. Moreover, the distinction between LoS and NLoS ranging mode can only be done in a stochastic manner with only some level of certainty. We discuss the use of the Schmidt Kalman filter as an algorithmic bias compensation method and propose to augment it with a novel constrained sigma point-based filtering method. On the other hand, to take into account the probabilistic nature of power-based NLoS identification, we develop an adaptive cooperative localization method with algorithmic bias compensation that seamlessly processes LoS and NLoS inter-agent measurements. The method has the advantage of being easy to be applied as an augmentation server atop local filters and becoming active only when an inter-agent range measurement is obtained. The low computational cost of the adaptive cooperative localization with the algorithmic bias compensation method enables real-time implementations. For UWB wireless communication, the current UWB communication protocols due to the half-duplex nature suffer from packet loss during data transmission in a cooperative localization setting. We develop a negotiation-based rescheduling dynamic time division multiple access (TDMA) protocol that has packet collision avoidance to achieve effective and energy-efficient UWB communication. In short, this dissertation work provides an effective UWB based cooperative localization solution that operates in real-time in a complex indoor environment. Field tests via a portable localization unit that is developed as part of this dissertation work demonstrate the effectiveness of our proposed solutions.